Postgres monitoring fails to reconnect when postgres is restarted

[evanj]

(I apologize if this is the wrong way to report agent bugs; feel free to point me to the correct location)

I'm running datadog-agent using the official Ubuntu package (4.0.1-236). Today I ran into the following problem:

I'm using the built-in Postgres monitoring connection, configured using /etc/dd-agent/conf.d/postgres.yaml to connect to localhost. Today, I restarted Postgres for an upgrade. A few hours later, I noticed that datadog agent wasn't collecting statistics. The following errors appeared in syslog. A restart of datadog agent fixed the problem.

I'm guessing datadog needs to detect when the connection is broken and attempt to re-connect?

Jan 3 15:58:15 ip-10-99-19-84 dd.collector[22640]: ERROR (init.py:468): Check 'postgres' instance #0 failed#012Traceback (most recent call last):#12 File "/usr/share/datadog/agent/checks/init.py", line 459, in run#012 self.check(instance)#12 File "/usr/share/datadog/agent/checks.d/postgres.py", line 152, in check#012 self._collect_stats(key, db, tags)#12 File "/usr/share/datadog/agent/checks.d/postgres.py", line 80, in _collect_stats#012 cursor = db.cursor()#012InterfaceError: connection already closed

[remh]

Thanks @evanj
This behavior will be fixed with the next version of the agent!